Re: [dhcwg] Bulk Leasequery for DHCPv6

["Marcus Goller" <mgoller at gmail.com>]

Hi Mark,

I agree that allowing the UDP based queries on TCP should be possible, the other way round might have more of an impact.

Another option, in my opinion, would be to introduce an new status code ("UnsupportedTransportLayer" or something similar) to indicate to the requestor that another protocol has to be used for this query type. That would also allow the DHCP server vendors to decide whether to allow UDP based query types on TCP, if not ruled out by your draft.

Regards,

Marcus

On 10/16/07, Mark Stapp <mjs at cisco.com> wrote:

Hello Marcus,

thanks for the rapid response!

a sketch of the history sort of goes like this: the early v6 leasequery
drafts did include semantics allowing for queries that could return
multiple/many results. during the WG discussion, there was some feeling
that trying to do that in UDP was making the leasequery protocol overly
complicated, trying to deal with marshalling, retransmission, etc. and
so the 'bulk' capability was taken out of what has become RFC5007. [if
I'm misrepresenting this, I hope someone will correct me.]

but there are some situations (that I try to describe in the new draft)
where a 'bulk' or 'one query, many results' capability may be very
useful. so I've tried to work out what a TCP-based protocol would look
like, mainly trying to extend the concepts that were introduced in RFC5007.

I do agree that one of the weaknesses in this first version of this
'bulk' draft is that there's no really tight language about the
relationship between the new queries and the existing UDP protocol. I
hope that that'll improve with WG input, as folks help decide how to
draw the lines between the two versions. personally, I think that
allowing the UDP queries in the TCP version should be fine. but there
isn't really any good way to deliver thousands of results in the UDP
version, so I don't think it's going to be worth much to try to permit
the use of 'bulk' queries in UDP.

Thanks,
Mark

Marcus Goller wrote:
> Hi,
>
> After a first quick read through, this sounds like a useful extension to
> the existing Leasequery mechanism. Would you mind quickly summarizing
> the reasons why supporting this with UDP queries would make the protocol
> too complicated?
> I agree that TCP is probably the better choice, but I would at least
> like to see UDP as an optional transport mechanism. The reason for this
> is that I am not convinced that protocol specific options (query types)
> should be dependent on the underlying transport protocol. The requestor
> should be able to choose the appropriate protocol. But I might
> underestimate the impact to the existing UDP based Leasequery
> specification, so any additional information would be appreciated.
>
> Regards,
>
> Marcus
>
>

_______________________________________________
dhcwg mailing list
dhcwg at ietf.org
https://www1.ietf.org/mailman/listinfo/dhcwg